SlideShare ist ein Scribd-Unternehmen logo
1 von 12
Downloaden Sie, um offline zu lesen
Graph Transplant: Node Saliency-Guided
Graph Mixup with Local Structure
Preservation
Joonhyung Park ∗, Hajin Shim ∗, Eunho Yang
AAAI 2022
Machine Learning & Intelligence Laboratory
∗
: Equal Contribution
• Generating virtual augmented data via interpolation of two images
Mixup-based Augmentation
Image 𝐗𝐀
Image 𝐗𝐁
CutMix (ICCV’19)
Mixup (ICLR’18)
𝑌
𝐴 (Fish)
0.5
𝑌𝐵 (Bird)
0.5
0.85
0.15
𝑌
𝐴 (Fish) 𝑌𝐵 (Bird)
• Advantages of Mixup-based augmentations
• Improve generalization performance (smooth decision boundary) 1
• Refine model calibration and predictive uncertainty of neural networks 2
• Enhance model robustness to data corruption 3
Mixup-based Augmentation
1) Manifold Mixup: Better Representations by Interpolating Hidden States, ICML’19
2) On Mixup Training: Improved Calibration and Predictive Uncertainty for Deep Neural Networks, NeurIPS’19
3) AugMix: A Simple Data Processing Method to Improve Robustness and Uncertainty, ICLR’20
• Non-trivial challenges for Graph Mixup
• No node correspondence
• The number of nodes may differ across graph instances
• Irregular connectivity
How about Graph-structured Data?
We propose the first Mixup-like graph augmentation method Graph Transplant
Graph 𝑮𝝅 Graph 𝑮
How to interpolate two dissimilar-structured graphs?
• 𝒌-hop subgraph is defined as a mixing unit that can preserve the local information
• Graph Transplant employs the node saliency information to select meaningful
subgraphs
• Node saliency encodes the importance of each node in classifying graph property
Graph Transplant: Salient Subgraph Extraction
Saliency of node 𝒗
Salient nodes
• 𝒌-hop subgraph is defined as a mixing unit that can preserve the local information
• Graph Transplant employs the node saliency information to select meaningful
subgraphs
• Salient subgraph is extracted in to be attached with another random subgraph
• Salient subgraph: 𝑘-hop subgraph anchored by salient nodes
Graph Transplant: Salient Subgraph Extraction
Graph 𝐺𝜋
Graph 𝐺
Salient Subgraph of 𝐺𝜋
Random Subgraph of 𝐺𝜋
: Anchor nodes
Graph Transplant: Connecting Two Subgraphs
• After extracting subgraphs, subgraphs are transplanted by adding new edges
• We propose two approaches for determining the connectivity of subgraphs
• DP (Degree Preservation): Random connection under the constraint on the original node
degree (preserves the expected degree of the nodes in the original graphs)
• EP (Edge Prediction): Predict edge probability by using a differentiable edge predictor
based on node features
Concept of Edge Prediction
Graph Transplant: Adaptive Label Mixing
• Assigning the appropriate label to the generated graph is necessary to prevent the
network from being misled by inappropriate supervision
• Simply assigning the label in proportion to the number of nodes will NOT explain the
generated graph properly
• Our label mixing strategy determines the label for mixed graph as the ratio of the
saliency between subgraphs
• The label ℓ for the mixed graph 𝐺′
is defined as:
∗ ℐ 𝑆 , ഥ
𝑉 =
σ𝑣∈𝑉 𝑠𝑣
𝑆
(Importance function; total saliency ratio of the nodes constituting the subgraph in the full graph)
ℓ 𝑦𝐺𝜋
, 𝑦𝐺 = 𝜆𝐺′𝑦𝐺𝜋
+ 1 − 𝜆𝐺′ 𝑦𝐺
𝜆𝐺′ =
ℐ(𝑆𝐺𝜋
, ഥ
𝑉
𝜋 )
ℐ 𝑆𝐺𝜋
, ഥ
𝑉
𝜋 + ℐ(𝑆𝐺 , 𝑉ത
𝑉)
Importance of Subgraph derived from 𝐺𝜋
Importance of Subgraph derived from 𝐺
• Graph Transplant consistently achieves the best performances for 5 graph classification
benchmark datasets
• Graph Transplant consistently exhibits superiority for datasets with different
characteristics such as sparsity of connection, graph size, and the presence or absence
of node features
Graph Classification Results
• We test if the GNNs trained with Graph Transplant can enhance the model robustness
against adversarial attacks.
• GradArgmax (Dai et al. 2018): Gradient based white-box attack designed for graph-
structured data
• Interestingly, although the adversarial attack is based on edge modification, ours
exhibits better performance than PermE, an edge perturbation augmentation.
Robustness to Adversarial Attack
• We analyze the node saliency of the molecular graphs to show that our method leads
the model to attend where it really matters to correctly classify molecules
• The task is to decide whether the molecule causes DNA mutation or not, which is
important to capture the interaction of functional groups
• E.g. The left case in Figure is of a mutagenic molecule in that it has two methyl groups attached to oxygen atoms,
which are vulnerable to radical attack. Attacked by radical, the attached methyl groups are transferred to DNA
base, entailing the change of DNA property, resulting in mutation. While the model trained with Graph
Transplant attends to the appropriate region, the vanilla model gives an incorrect answer with highlighting the
unrelated spot
Qualitative Analysis
• We proposed Graph Transplant, the first input-level Mixup strategy for graph data, that
generates mixed graphs via the guidance of the node saliency
• Our contribution
• We propose Graph Transplant, that can mix two dissimilar-structured graphs by replacing the
destination subgraph with the source subgraph while preserving the local structure
• In contrast to random augmentations, we present a novel strategy to generate properly mixed
graphs with adaptively assigned labels that adheres to the corresponding labels by explicitly
leveraging the node saliency
• Through extensive experiments, we show that Graph Transplant is an effective algorithm that
can bring overall improvement across the multiple fields of graphs in terms of the
classification performance, robustness, and model calibration.
Conclusion

Weitere ähnliche Inhalte

Was ist angesagt?

Semi-Targeted Model Poisoning Attack on Federated Learning via Backward Error...
Semi-Targeted Model Poisoning Attack on Federated Learning via Backward Error...Semi-Targeted Model Poisoning Attack on Federated Learning via Backward Error...
Semi-Targeted Model Poisoning Attack on Federated Learning via Backward Error...
YuweiSun5
 
Data mining to predict academic performance.
Data mining to predict academic performance. Data mining to predict academic performance.
Data mining to predict academic performance.
Ranjith Gowda
 

Was ist angesagt? (20)

Semi-Targeted Model Poisoning Attack on Federated Learning via Backward Error...
Semi-Targeted Model Poisoning Attack on Federated Learning via Backward Error...Semi-Targeted Model Poisoning Attack on Federated Learning via Backward Error...
Semi-Targeted Model Poisoning Attack on Federated Learning via Backward Error...
 
Risk Signature Profiles in Health Care Claims(Risk_Signature_Profiles)_.pptx
Risk Signature Profiles in Health Care Claims(Risk_Signature_Profiles)_.pptxRisk Signature Profiles in Health Care Claims(Risk_Signature_Profiles)_.pptx
Risk Signature Profiles in Health Care Claims(Risk_Signature_Profiles)_.pptx
 
A friendly introduction to GANs
A friendly introduction to GANsA friendly introduction to GANs
A friendly introduction to GANs
 
Tutorial on Generalization in Neural Fields, CVPR 2022 Tutorial on Neural Fie...
Tutorial on Generalization in Neural Fields, CVPR 2022 Tutorial on Neural Fie...Tutorial on Generalization in Neural Fields, CVPR 2022 Tutorial on Neural Fie...
Tutorial on Generalization in Neural Fields, CVPR 2022 Tutorial on Neural Fie...
 
CUDAメモ
CUDAメモCUDAメモ
CUDAメモ
 
Neo4j GraphSummit London - The Path To Success With Graph Database and Data S...
Neo4j GraphSummit London - The Path To Success With Graph Database and Data S...Neo4j GraphSummit London - The Path To Success With Graph Database and Data S...
Neo4j GraphSummit London - The Path To Success With Graph Database and Data S...
 
Off-line Signature Verification
Off-line Signature VerificationOff-line Signature Verification
Off-line Signature Verification
 
[DL輪読会]Domain Adaptive Faster R-CNN for Object Detection in the Wild
[DL輪読会]Domain Adaptive Faster R-CNN for Object Detection in the Wild[DL輪読会]Domain Adaptive Faster R-CNN for Object Detection in the Wild
[DL輪読会]Domain Adaptive Faster R-CNN for Object Detection in the Wild
 
Network embedding
Network embeddingNetwork embedding
Network embedding
 
スマートフォンカメラとペットボトルを用いた握力計測手法の提案(SIG-DeMO研究会)
スマートフォンカメラとペットボトルを用いた握力計測手法の提案(SIG-DeMO研究会)スマートフォンカメラとペットボトルを用いた握力計測手法の提案(SIG-DeMO研究会)
スマートフォンカメラとペットボトルを用いた握力計測手法の提案(SIG-DeMO研究会)
 
Proximal Policy Optimization (Reinforcement Learning)
Proximal Policy Optimization (Reinforcement Learning)Proximal Policy Optimization (Reinforcement Learning)
Proximal Policy Optimization (Reinforcement Learning)
 
GraphTour 2020 - Danish Business Authority: First line of Defence
GraphTour 2020 - Danish Business Authority: First line of DefenceGraphTour 2020 - Danish Business Authority: First line of Defence
GraphTour 2020 - Danish Business Authority: First line of Defence
 
DeepWalk: Online Learning of Representations
DeepWalk: Online Learning of RepresentationsDeepWalk: Online Learning of Representations
DeepWalk: Online Learning of Representations
 
How Powerful are Graph Networks?
How Powerful are Graph Networks?How Powerful are Graph Networks?
How Powerful are Graph Networks?
 
Data mining to predict academic performance.
Data mining to predict academic performance. Data mining to predict academic performance.
Data mining to predict academic performance.
 
Deepwalk vs Node2vec
Deepwalk vs Node2vecDeepwalk vs Node2vec
Deepwalk vs Node2vec
 
CVPR 2018 速報
CVPR 2018 速報CVPR 2018 速報
CVPR 2018 速報
 
Biometric Security advantages and disadvantages
Biometric Security advantages and disadvantagesBiometric Security advantages and disadvantages
Biometric Security advantages and disadvantages
 
Mask-RCNNを用いたキャベツの結球認識
Mask-RCNNを用いたキャベツの結球認識Mask-RCNNを用いたキャベツの結球認識
Mask-RCNNを用いたキャベツの結球認識
 
Weisfeiler and Leman Go Neural: Higher-order Graph Neural Networks
Weisfeiler and Leman Go Neural: Higher-order Graph Neural Networks Weisfeiler and Leman Go Neural: Higher-order Graph Neural Networks
Weisfeiler and Leman Go Neural: Higher-order Graph Neural Networks
 

Ähnlich wie J. Park, H. Shim, AAAI 2022, MLILAB, KAISTAI

Edge Representation Learning with Hypergraphs
Edge Representation Learning with HypergraphsEdge Representation Learning with Hypergraphs
Edge Representation Learning with Hypergraphs
MLAI2
 
Dahlquist et-al bosc-ismb_2016_poster
Dahlquist et-al bosc-ismb_2016_posterDahlquist et-al bosc-ismb_2016_poster
Dahlquist et-al bosc-ismb_2016_poster
GRNsight
 
IJCER (www.ijceronline.com) International Journal of computational Engineerin...
IJCER (www.ijceronline.com) International Journal of computational Engineerin...IJCER (www.ijceronline.com) International Journal of computational Engineerin...
IJCER (www.ijceronline.com) International Journal of computational Engineerin...
ijceronline
 
various methods for image segmentation
various methods for image segmentationvarious methods for image segmentation
various methods for image segmentation
Raveesh Methi
 
A simple framework for contrastive learning of visual representations
A simple framework for contrastive learning of visual representationsA simple framework for contrastive learning of visual representations
A simple framework for contrastive learning of visual representations
Devansh16
 
Classification of handwritten characters by their symmetry features
Classification of handwritten characters by their symmetry featuresClassification of handwritten characters by their symmetry features
Classification of handwritten characters by their symmetry features
AYUSH RAJ
 
A STUDY AND ANALYSIS OF DIFFERENT EDGE DETECTION TECHNIQUES
A STUDY AND ANALYSIS OF DIFFERENT EDGE DETECTION TECHNIQUESA STUDY AND ANALYSIS OF DIFFERENT EDGE DETECTION TECHNIQUES
A STUDY AND ANALYSIS OF DIFFERENT EDGE DETECTION TECHNIQUES
cscpconf
 

Ähnlich wie J. Park, H. Shim, AAAI 2022, MLILAB, KAISTAI (20)

NS-CUK Joint Journal Club : S.T.Nguyen, Review on "Graph Neural Networks for ...
NS-CUK Joint Journal Club : S.T.Nguyen, Review on "Graph Neural Networks for ...NS-CUK Joint Journal Club : S.T.Nguyen, Review on "Graph Neural Networks for ...
NS-CUK Joint Journal Club : S.T.Nguyen, Review on "Graph Neural Networks for ...
 
Edge Representation Learning with Hypergraphs
Edge Representation Learning with HypergraphsEdge Representation Learning with Hypergraphs
Edge Representation Learning with Hypergraphs
 
Dahlquist et-al bosc-ismb_2016_poster
Dahlquist et-al bosc-ismb_2016_posterDahlquist et-al bosc-ismb_2016_poster
Dahlquist et-al bosc-ismb_2016_poster
 
NS-CUK Seminar: S.T.Nguyen, Review on "Improving Graph Neural Network Express...
NS-CUK Seminar: S.T.Nguyen, Review on "Improving Graph Neural Network Express...NS-CUK Seminar: S.T.Nguyen, Review on "Improving Graph Neural Network Express...
NS-CUK Seminar: S.T.Nguyen, Review on "Improving Graph Neural Network Express...
 
IJCER (www.ijceronline.com) International Journal of computational Engineerin...
IJCER (www.ijceronline.com) International Journal of computational Engineerin...IJCER (www.ijceronline.com) International Journal of computational Engineerin...
IJCER (www.ijceronline.com) International Journal of computational Engineerin...
 
Chapter 4 better.pptx
Chapter 4 better.pptxChapter 4 better.pptx
Chapter 4 better.pptx
 
various methods for image segmentation
various methods for image segmentationvarious methods for image segmentation
various methods for image segmentation
 
Stereo Matching by Deep Learning
Stereo Matching by Deep LearningStereo Matching by Deep Learning
Stereo Matching by Deep Learning
 
Markov Cluster Algorithm & real world application
Markov Cluster Algorithm & real world applicationMarkov Cluster Algorithm & real world application
Markov Cluster Algorithm & real world application
 
Multiple Ant Colony Optimizations for Stereo Matching
Multiple Ant Colony Optimizations for Stereo MatchingMultiple Ant Colony Optimizations for Stereo Matching
Multiple Ant Colony Optimizations for Stereo Matching
 
Effective Pixel Interpolation for Image Super Resolution
Effective Pixel Interpolation for Image Super ResolutionEffective Pixel Interpolation for Image Super Resolution
Effective Pixel Interpolation for Image Super Resolution
 
Effective Pixel Interpolation for Image Super Resolution
Effective Pixel Interpolation for Image Super ResolutionEffective Pixel Interpolation for Image Super Resolution
Effective Pixel Interpolation for Image Super Resolution
 
Exploring Simple Siamese Representation Learning
Exploring Simple Siamese Representation LearningExploring Simple Siamese Representation Learning
Exploring Simple Siamese Representation Learning
 
A simple framework for contrastive learning of visual representations
A simple framework for contrastive learning of visual representationsA simple framework for contrastive learning of visual representations
A simple framework for contrastive learning of visual representations
 
2006 ssiai
2006 ssiai2006 ssiai
2006 ssiai
 
"Sparse Graph Attention Networks", IEEE Transactions on Knowledge and Data En...
"Sparse Graph Attention Networks", IEEE Transactions on Knowledge and Data En..."Sparse Graph Attention Networks", IEEE Transactions on Knowledge and Data En...
"Sparse Graph Attention Networks", IEEE Transactions on Knowledge and Data En...
 
Classification of handwritten characters by their symmetry features
Classification of handwritten characters by their symmetry featuresClassification of handwritten characters by their symmetry features
Classification of handwritten characters by their symmetry features
 
A STUDY AND ANALYSIS OF DIFFERENT EDGE DETECTION TECHNIQUES
A STUDY AND ANALYSIS OF DIFFERENT EDGE DETECTION TECHNIQUESA STUDY AND ANALYSIS OF DIFFERENT EDGE DETECTION TECHNIQUES
A STUDY AND ANALYSIS OF DIFFERENT EDGE DETECTION TECHNIQUES
 
Decomposing image generation into layout priction and conditional synthesis
Decomposing image generation into layout priction and conditional synthesisDecomposing image generation into layout priction and conditional synthesis
Decomposing image generation into layout priction and conditional synthesis
 
Sun_MAPL_GNN.pptx
Sun_MAPL_GNN.pptxSun_MAPL_GNN.pptx
Sun_MAPL_GNN.pptx
 

Mehr von MLILAB

H. Shim, NeurIPS 2018, MLILAB, KAIST AI
H. Shim, NeurIPS 2018, MLILAB, KAIST AIH. Shim, NeurIPS 2018, MLILAB, KAIST AI
H. Shim, NeurIPS 2018, MLILAB, KAIST AI
MLILAB
 

Mehr von MLILAB (20)

J. Jeong, AAAI 2024, MLILAB, KAIST AI..
J. Jeong,  AAAI 2024, MLILAB, KAIST AI..J. Jeong,  AAAI 2024, MLILAB, KAIST AI..
J. Jeong, AAAI 2024, MLILAB, KAIST AI..
 
J. Yun, NeurIPS 2023, MLILAB, KAISTAI
J. Yun,  NeurIPS 2023,  MLILAB,  KAISTAIJ. Yun,  NeurIPS 2023,  MLILAB,  KAISTAI
J. Yun, NeurIPS 2023, MLILAB, KAISTAI
 
S. Kim, NeurIPS 2023, MLILAB, KAISTAI
S. Kim,  NeurIPS 2023,  MLILAB,  KAISTAIS. Kim,  NeurIPS 2023,  MLILAB,  KAISTAI
S. Kim, NeurIPS 2023, MLILAB, KAISTAI
 
C. Kim, INTERSPEECH 2023, MLILAB, KAISTAI
C. Kim, INTERSPEECH 2023, MLILAB, KAISTAIC. Kim, INTERSPEECH 2023, MLILAB, KAISTAI
C. Kim, INTERSPEECH 2023, MLILAB, KAISTAI
 
Y. Jung, ICML 2023, MLILAB, KAISTAI
Y. Jung, ICML 2023, MLILAB, KAISTAIY. Jung, ICML 2023, MLILAB, KAISTAI
Y. Jung, ICML 2023, MLILAB, KAISTAI
 
J. Song, S. Kim, ICML 2023, MLILAB, KAISTAI
J. Song, S. Kim, ICML 2023, MLILAB, KAISTAIJ. Song, S. Kim, ICML 2023, MLILAB, KAISTAI
J. Song, S. Kim, ICML 2023, MLILAB, KAISTAI
 
K. Seo, ICASSP 2023, MLILAB, KAISTAI
K. Seo, ICASSP 2023, MLILAB, KAISTAIK. Seo, ICASSP 2023, MLILAB, KAISTAI
K. Seo, ICASSP 2023, MLILAB, KAISTAI
 
G. Kim, CVPR 2023, MLILAB, KAISTAI
G. Kim, CVPR 2023, MLILAB, KAISTAIG. Kim, CVPR 2023, MLILAB, KAISTAI
G. Kim, CVPR 2023, MLILAB, KAISTAI
 
S. Kim, ICLR 2023, MLILAB, KAISTAI
S. Kim, ICLR 2023, MLILAB, KAISTAIS. Kim, ICLR 2023, MLILAB, KAISTAI
S. Kim, ICLR 2023, MLILAB, KAISTAI
 
Y. Kim, ICLR 2023, MLILAB, KAISTAI
Y. Kim, ICLR 2023, MLILAB, KAISTAIY. Kim, ICLR 2023, MLILAB, KAISTAI
Y. Kim, ICLR 2023, MLILAB, KAISTAI
 
J. Yun, AISTATS 2022, MLILAB, KAISTAI
J. Yun, AISTATS 2022, MLILAB, KAISTAIJ. Yun, AISTATS 2022, MLILAB, KAISTAI
J. Yun, AISTATS 2022, MLILAB, KAISTAI
 
J. Song, J. Park, ICML 2022, MLILAB, KAISTAI
J. Song, J. Park, ICML 2022, MLILAB, KAISTAIJ. Song, J. Park, ICML 2022, MLILAB, KAISTAI
J. Song, J. Park, ICML 2022, MLILAB, KAISTAI
 
J. Park, J. Song, ICLR 2022, MLILAB, KAISTAI
J. Park, J. Song, ICLR 2022, MLILAB, KAISTAIJ. Park, J. Song, ICLR 2022, MLILAB, KAISTAI
J. Park, J. Song, ICLR 2022, MLILAB, KAISTAI
 
J. Park, AAAI 2022, MLILAB, KAIST AI
J. Park, AAAI 2022, MLILAB, KAIST AIJ. Park, AAAI 2022, MLILAB, KAIST AI
J. Park, AAAI 2022, MLILAB, KAIST AI
 
J. Song, et. al., ASRU 2021, MLILAB, KAIST AI
J. Song, et. al., ASRU 2021, MLILAB, KAIST AIJ. Song, et. al., ASRU 2021, MLILAB, KAIST AI
J. Song, et. al., ASRU 2021, MLILAB, KAIST AI
 
J. Song, H. Shim et al., ICASSP 2021, MLILAB, KAIST AI
J. Song, H. Shim et al., ICASSP 2021, MLILAB, KAIST AIJ. Song, H. Shim et al., ICASSP 2021, MLILAB, KAIST AI
J. Song, H. Shim et al., ICASSP 2021, MLILAB, KAIST AI
 
T. Yoon, et. al., ICLR 2021, MLILAB, KAIST AI
T. Yoon, et. al., ICLR 2021, MLILAB, KAIST AIT. Yoon, et. al., ICLR 2021, MLILAB, KAIST AI
T. Yoon, et. al., ICLR 2021, MLILAB, KAIST AI
 
G. Park, J.-Y. Yang, et. al., NeurIPS 2020, MLILAB, KAIST AI
G. Park, J.-Y. Yang, et. al., NeurIPS 2020, MLILAB, KAIST AIG. Park, J.-Y. Yang, et. al., NeurIPS 2020, MLILAB, KAIST AI
G. Park, J.-Y. Yang, et. al., NeurIPS 2020, MLILAB, KAIST AI
 
I. Chung, AAAI 2020, MLILAB, KAIST AI
I. Chung, AAAI 2020, MLILAB, KAIST AII. Chung, AAAI 2020, MLILAB, KAIST AI
I. Chung, AAAI 2020, MLILAB, KAIST AI
 
H. Shim, NeurIPS 2018, MLILAB, KAIST AI
H. Shim, NeurIPS 2018, MLILAB, KAIST AIH. Shim, NeurIPS 2018, MLILAB, KAIST AI
H. Shim, NeurIPS 2018, MLILAB, KAIST AI
 

Kürzlich hochgeladen

Structural Analysis and Design of Foundations: A Comprehensive Handbook for S...
Structural Analysis and Design of Foundations: A Comprehensive Handbook for S...Structural Analysis and Design of Foundations: A Comprehensive Handbook for S...
Structural Analysis and Design of Foundations: A Comprehensive Handbook for S...
Dr.Costas Sachpazis
 
Call Now ≽ 9953056974 ≼🔝 Call Girls In New Ashok Nagar ≼🔝 Delhi door step de...
Call Now ≽ 9953056974 ≼🔝 Call Girls In New Ashok Nagar  ≼🔝 Delhi door step de...Call Now ≽ 9953056974 ≼🔝 Call Girls In New Ashok Nagar  ≼🔝 Delhi door step de...
Call Now ≽ 9953056974 ≼🔝 Call Girls In New Ashok Nagar ≼🔝 Delhi door step de...
9953056974 Low Rate Call Girls In Saket, Delhi NCR
 
Call for Papers - Educational Administration: Theory and Practice, E-ISSN: 21...
Call for Papers - Educational Administration: Theory and Practice, E-ISSN: 21...Call for Papers - Educational Administration: Theory and Practice, E-ISSN: 21...
Call for Papers - Educational Administration: Theory and Practice, E-ISSN: 21...
Christo Ananth
 
AKTU Computer Networks notes --- Unit 3.pdf
AKTU Computer Networks notes ---  Unit 3.pdfAKTU Computer Networks notes ---  Unit 3.pdf
AKTU Computer Networks notes --- Unit 3.pdf
ankushspencer015
 

Kürzlich hochgeladen (20)

Coefficient of Thermal Expansion and their Importance.pptx
Coefficient of Thermal Expansion and their Importance.pptxCoefficient of Thermal Expansion and their Importance.pptx
Coefficient of Thermal Expansion and their Importance.pptx
 
Booking open Available Pune Call Girls Koregaon Park 6297143586 Call Hot Ind...
Booking open Available Pune Call Girls Koregaon Park  6297143586 Call Hot Ind...Booking open Available Pune Call Girls Koregaon Park  6297143586 Call Hot Ind...
Booking open Available Pune Call Girls Koregaon Park 6297143586 Call Hot Ind...
 
Structural Analysis and Design of Foundations: A Comprehensive Handbook for S...
Structural Analysis and Design of Foundations: A Comprehensive Handbook for S...Structural Analysis and Design of Foundations: A Comprehensive Handbook for S...
Structural Analysis and Design of Foundations: A Comprehensive Handbook for S...
 
Thermal Engineering -unit - III & IV.ppt
Thermal Engineering -unit - III & IV.pptThermal Engineering -unit - III & IV.ppt
Thermal Engineering -unit - III & IV.ppt
 
Call Now ≽ 9953056974 ≼🔝 Call Girls In New Ashok Nagar ≼🔝 Delhi door step de...
Call Now ≽ 9953056974 ≼🔝 Call Girls In New Ashok Nagar  ≼🔝 Delhi door step de...Call Now ≽ 9953056974 ≼🔝 Call Girls In New Ashok Nagar  ≼🔝 Delhi door step de...
Call Now ≽ 9953056974 ≼🔝 Call Girls In New Ashok Nagar ≼🔝 Delhi door step de...
 
Double rodded leveling 1 pdf activity 01
Double rodded leveling 1 pdf activity 01Double rodded leveling 1 pdf activity 01
Double rodded leveling 1 pdf activity 01
 
Call for Papers - Educational Administration: Theory and Practice, E-ISSN: 21...
Call for Papers - Educational Administration: Theory and Practice, E-ISSN: 21...Call for Papers - Educational Administration: Theory and Practice, E-ISSN: 21...
Call for Papers - Educational Administration: Theory and Practice, E-ISSN: 21...
 
UNIT - IV - Air Compressors and its Performance
UNIT - IV - Air Compressors and its PerformanceUNIT - IV - Air Compressors and its Performance
UNIT - IV - Air Compressors and its Performance
 
data_management_and _data_science_cheat_sheet.pdf
data_management_and _data_science_cheat_sheet.pdfdata_management_and _data_science_cheat_sheet.pdf
data_management_and _data_science_cheat_sheet.pdf
 
chapter 5.pptx: drainage and irrigation engineering
chapter 5.pptx: drainage and irrigation engineeringchapter 5.pptx: drainage and irrigation engineering
chapter 5.pptx: drainage and irrigation engineering
 
AKTU Computer Networks notes --- Unit 3.pdf
AKTU Computer Networks notes ---  Unit 3.pdfAKTU Computer Networks notes ---  Unit 3.pdf
AKTU Computer Networks notes --- Unit 3.pdf
 
UNIT-IFLUID PROPERTIES & FLOW CHARACTERISTICS
UNIT-IFLUID PROPERTIES & FLOW CHARACTERISTICSUNIT-IFLUID PROPERTIES & FLOW CHARACTERISTICS
UNIT-IFLUID PROPERTIES & FLOW CHARACTERISTICS
 
Call for Papers - International Journal of Intelligent Systems and Applicatio...
Call for Papers - International Journal of Intelligent Systems and Applicatio...Call for Papers - International Journal of Intelligent Systems and Applicatio...
Call for Papers - International Journal of Intelligent Systems and Applicatio...
 
VIP Model Call Girls Kothrud ( Pune ) Call ON 8005736733 Starting From 5K to ...
VIP Model Call Girls Kothrud ( Pune ) Call ON 8005736733 Starting From 5K to ...VIP Model Call Girls Kothrud ( Pune ) Call ON 8005736733 Starting From 5K to ...
VIP Model Call Girls Kothrud ( Pune ) Call ON 8005736733 Starting From 5K to ...
 
The Most Attractive Pune Call Girls Budhwar Peth 8250192130 Will You Miss Thi...
The Most Attractive Pune Call Girls Budhwar Peth 8250192130 Will You Miss Thi...The Most Attractive Pune Call Girls Budhwar Peth 8250192130 Will You Miss Thi...
The Most Attractive Pune Call Girls Budhwar Peth 8250192130 Will You Miss Thi...
 
Top Rated Pune Call Girls Budhwar Peth ⟟ 6297143586 ⟟ Call Me For Genuine Se...
Top Rated  Pune Call Girls Budhwar Peth ⟟ 6297143586 ⟟ Call Me For Genuine Se...Top Rated  Pune Call Girls Budhwar Peth ⟟ 6297143586 ⟟ Call Me For Genuine Se...
Top Rated Pune Call Girls Budhwar Peth ⟟ 6297143586 ⟟ Call Me For Genuine Se...
 
Booking open Available Pune Call Girls Pargaon 6297143586 Call Hot Indian Gi...
Booking open Available Pune Call Girls Pargaon  6297143586 Call Hot Indian Gi...Booking open Available Pune Call Girls Pargaon  6297143586 Call Hot Indian Gi...
Booking open Available Pune Call Girls Pargaon 6297143586 Call Hot Indian Gi...
 
KubeKraft presentation @CloudNativeHooghly
KubeKraft presentation @CloudNativeHooghlyKubeKraft presentation @CloudNativeHooghly
KubeKraft presentation @CloudNativeHooghly
 
Thermal Engineering Unit - I & II . ppt
Thermal Engineering  Unit - I & II . pptThermal Engineering  Unit - I & II . ppt
Thermal Engineering Unit - I & II . ppt
 
(INDIRA) Call Girl Bhosari Call Now 8617697112 Bhosari Escorts 24x7
(INDIRA) Call Girl Bhosari Call Now 8617697112 Bhosari Escorts 24x7(INDIRA) Call Girl Bhosari Call Now 8617697112 Bhosari Escorts 24x7
(INDIRA) Call Girl Bhosari Call Now 8617697112 Bhosari Escorts 24x7
 

J. Park, H. Shim, AAAI 2022, MLILAB, KAISTAI

  • 1. Graph Transplant: Node Saliency-Guided Graph Mixup with Local Structure Preservation Joonhyung Park ∗, Hajin Shim ∗, Eunho Yang AAAI 2022 Machine Learning & Intelligence Laboratory ∗ : Equal Contribution
  • 2. • Generating virtual augmented data via interpolation of two images Mixup-based Augmentation Image 𝐗𝐀 Image 𝐗𝐁 CutMix (ICCV’19) Mixup (ICLR’18) 𝑌 𝐴 (Fish) 0.5 𝑌𝐵 (Bird) 0.5 0.85 0.15 𝑌 𝐴 (Fish) 𝑌𝐵 (Bird)
  • 3. • Advantages of Mixup-based augmentations • Improve generalization performance (smooth decision boundary) 1 • Refine model calibration and predictive uncertainty of neural networks 2 • Enhance model robustness to data corruption 3 Mixup-based Augmentation 1) Manifold Mixup: Better Representations by Interpolating Hidden States, ICML’19 2) On Mixup Training: Improved Calibration and Predictive Uncertainty for Deep Neural Networks, NeurIPS’19 3) AugMix: A Simple Data Processing Method to Improve Robustness and Uncertainty, ICLR’20
  • 4. • Non-trivial challenges for Graph Mixup • No node correspondence • The number of nodes may differ across graph instances • Irregular connectivity How about Graph-structured Data? We propose the first Mixup-like graph augmentation method Graph Transplant Graph 𝑮𝝅 Graph 𝑮 How to interpolate two dissimilar-structured graphs?
  • 5. • 𝒌-hop subgraph is defined as a mixing unit that can preserve the local information • Graph Transplant employs the node saliency information to select meaningful subgraphs • Node saliency encodes the importance of each node in classifying graph property Graph Transplant: Salient Subgraph Extraction Saliency of node 𝒗 Salient nodes
  • 6. • 𝒌-hop subgraph is defined as a mixing unit that can preserve the local information • Graph Transplant employs the node saliency information to select meaningful subgraphs • Salient subgraph is extracted in to be attached with another random subgraph • Salient subgraph: 𝑘-hop subgraph anchored by salient nodes Graph Transplant: Salient Subgraph Extraction Graph 𝐺𝜋 Graph 𝐺 Salient Subgraph of 𝐺𝜋 Random Subgraph of 𝐺𝜋 : Anchor nodes
  • 7. Graph Transplant: Connecting Two Subgraphs • After extracting subgraphs, subgraphs are transplanted by adding new edges • We propose two approaches for determining the connectivity of subgraphs • DP (Degree Preservation): Random connection under the constraint on the original node degree (preserves the expected degree of the nodes in the original graphs) • EP (Edge Prediction): Predict edge probability by using a differentiable edge predictor based on node features Concept of Edge Prediction
  • 8. Graph Transplant: Adaptive Label Mixing • Assigning the appropriate label to the generated graph is necessary to prevent the network from being misled by inappropriate supervision • Simply assigning the label in proportion to the number of nodes will NOT explain the generated graph properly • Our label mixing strategy determines the label for mixed graph as the ratio of the saliency between subgraphs • The label ℓ for the mixed graph 𝐺′ is defined as: ∗ ℐ 𝑆 , ഥ 𝑉 = σ𝑣∈𝑉 𝑠𝑣 𝑆 (Importance function; total saliency ratio of the nodes constituting the subgraph in the full graph) ℓ 𝑦𝐺𝜋 , 𝑦𝐺 = 𝜆𝐺′𝑦𝐺𝜋 + 1 − 𝜆𝐺′ 𝑦𝐺 𝜆𝐺′ = ℐ(𝑆𝐺𝜋 , ഥ 𝑉 𝜋 ) ℐ 𝑆𝐺𝜋 , ഥ 𝑉 𝜋 + ℐ(𝑆𝐺 , 𝑉ത 𝑉) Importance of Subgraph derived from 𝐺𝜋 Importance of Subgraph derived from 𝐺
  • 9. • Graph Transplant consistently achieves the best performances for 5 graph classification benchmark datasets • Graph Transplant consistently exhibits superiority for datasets with different characteristics such as sparsity of connection, graph size, and the presence or absence of node features Graph Classification Results
  • 10. • We test if the GNNs trained with Graph Transplant can enhance the model robustness against adversarial attacks. • GradArgmax (Dai et al. 2018): Gradient based white-box attack designed for graph- structured data • Interestingly, although the adversarial attack is based on edge modification, ours exhibits better performance than PermE, an edge perturbation augmentation. Robustness to Adversarial Attack
  • 11. • We analyze the node saliency of the molecular graphs to show that our method leads the model to attend where it really matters to correctly classify molecules • The task is to decide whether the molecule causes DNA mutation or not, which is important to capture the interaction of functional groups • E.g. The left case in Figure is of a mutagenic molecule in that it has two methyl groups attached to oxygen atoms, which are vulnerable to radical attack. Attacked by radical, the attached methyl groups are transferred to DNA base, entailing the change of DNA property, resulting in mutation. While the model trained with Graph Transplant attends to the appropriate region, the vanilla model gives an incorrect answer with highlighting the unrelated spot Qualitative Analysis
  • 12. • We proposed Graph Transplant, the first input-level Mixup strategy for graph data, that generates mixed graphs via the guidance of the node saliency • Our contribution • We propose Graph Transplant, that can mix two dissimilar-structured graphs by replacing the destination subgraph with the source subgraph while preserving the local structure • In contrast to random augmentations, we present a novel strategy to generate properly mixed graphs with adaptively assigned labels that adheres to the corresponding labels by explicitly leveraging the node saliency • Through extensive experiments, we show that Graph Transplant is an effective algorithm that can bring overall improvement across the multiple fields of graphs in terms of the classification performance, robustness, and model calibration. Conclusion